基于调和模型的降质图像超分辨率复原算法  

Super-Resolution Restoration Algorithm of Degraded Image Based on Harmonic Model

作　　者：王媛芳 王鸿 廖武忠 WANG Yuan-fang;WANG Hong;LIAO Wu-zhong(Software School,Chongqing Institute of Engineering,Chongqing 400056,China;Chongqing University of Technology,School of Vehicle Engineering,Chongqing 400054,China)

机构地区：[1]重庆工程学院软件学院,重庆400056 [2]重庆理工大学车辆工程学院,重庆400054

出　　处：《计算机仿真》2022年第6期183-186,268,共5页Computer Simulation

基　　金：基于图像特征信息的图像拼接算法研究(kjqn201901904)。

摘　　要：受光照、噪声、数字化设备性能等因素影响,致使图像分辨率较低、目标特征模糊,为此提出基于调和模型的降质图像超分辨率复原算法。采用权重函数与动态补偿矩阵方式搭建降质图像观测模型,并根据图像RGB值重新组合原始图像与邻域像素RGB值,得出重构图像;依据复原图像的各部分方差数值大小判定出图像边缘位置,为避免过度平滑,凭借自适应正则化修正边缘四角,最终利用离散算法与网络能量函数完成降质图像超分辨率复原运算。仿真结果表明,所提算法具有良好的复原性、收敛性能,并且信噪比极佳。Due to some factors such as illumination, noise, and the performance of digital equipment, the image resolution is always low and the target features are fuzzy. For this reason, an algorithm of super-resolution restoration for degraded images based on a harmonic model was proposed. Firstly, weight function and dynamic compensation matrix were used to build an observation model for the degraded image. According to the RGB values of the image, the original image and the RGB values of neighboring pixels were recombined to obtain a reconstructed image. Secondly, according to the variance value of each part in the restored image, the edge of the image was determined. In order to avoid the over-smoothness of the image, four corners of the edge should be corrected by adaptive regularization. Finally, the operation for super-resolution restoration of the degraded image was completed by a discrete algorithm and network energy function. Simulation results show that the proposed algorithm has good resilience, convergence performance, and an excellent signal-to-noise ratio.

关 键 词：调和模型 降质图像 超分辨率复原 模糊矩阵 正则化算子 神经网络 

分 类 号：TP392[自动化与计算机技术—计算机应用技术]